Version zero padding vs. postrelease suffix

[AMDmi3]

From libversion point of view, 1.4rc2 != 1.4.0rc2, while from human point of view it's probably the same thing. Not sure how to fix that yet.

[AMDmi3]

Rough idea would be to save intermediate non-equal comparison result, then try to eat zeroes until both arguments are in -a1 form. If that succeeds, return new result, otherwise return saved result.

[AMDmi3]

Not sure if this won't break ordering in specific cases though, as in #17.